It is widely acknowledged [SKA14] that the prediction of turbulent flow features in the presence of separation is one of the most significant challenges in fluid dynamics.
Low cost simulation methods (RANS or even wall-modeled LES) allow for an extensive exploration of the design space, but suffer from lower reliability especially for separated and secondary flows. Improving model reliability will therefore have a major impact on energy consumption, emission and noise of aircraft, cars, and ships due to significant improvements in design.
The financial consequences would likely reach billions of euros in savings of time-to-market and cost of the whole aircraft-design chain.
This proposal is associated with the EU-H2020 project HiFi-TURB which aims at developing new turbulence models, based on high-fidelity DNS data and their subsequent analysis by artificial intelligence and big data methodologies.
The objective of this proposal is to generate and exploit in situ DNS data for improving LES wall-modeling for flows featuring separation on a representative and challenging configuration. This configuration is the smooth backward facing step which is relevant for many industrial flows. It features a separation bubble of which the start and extent is highly dependent on the correct capture of turbulent momentum transfer upstream.